0

IndoorAtlas claims that it's able to acquire exact location signal through a device's magnetometer for geolocation purposes. How is this possible? Is this a viable alternative to GPS, which does not work indoors, for determining where a person is at a given time when my app updates a user's location? Will it be able to tell me this user is in the CVS in a shopping mall, for example, as opposed to the Walgreens next door? Or even can it tell me I'm in the cheese aisle of the grocery store?

2 Answers2

0

IndoorAtlas works by mapping unique magnetic data gathered by the device, and pairs that up with a floorplan that is pre-set to geo-coordinates, then marries that data on the backend to get fine-grained indoor position. All of this data for the location is gathered during a mapping process. After matching the store to known geo-coordinates, you follow a path around the store and gather the unique magnetic data for the floorplan.

Knowing the difference between the CVS or Walgreens would be a matter for the app developer. Currently, you have to select the floorplan you are in. Ideally, the retailer would locate you via GPS to get the right store, then provide the indoor map once you are there. The app shouldn't work in another store because the magnetic fingerprint is different. According to their FAQ, the ability to auto-select floorplans is on their future roadmap.

On a finer scale, the magnetic data in the cheese aisle would be unique, and IndoorAtlas would be able to distinguish that and return correct positioning data. The accuracy is about 3 meters.

Frackinfrell
  • 327
  • 2
  • 12
0

I am testing my own algorithm and so far I managed to turn my iPhone into a metal detector, only with the calibrated magnetometer data. You need measure the total strength of the magnetic field and essentially measure the total strength in several points within the same area that you want to "fingerprint". There are some academic papers written by people that have gone further than I have and essentially use the teqnique broadly described above. Here is a very good example where several meters accuracy seem doable:

http://www.mdpi.com/1424-8220/14/6/11001/pdf

Eppilo
  • 763
  • 6
  • 19